Exploring the specificity and predictive ability of multiple domains of spatial ability on STEM educational achievement, career choice and STEM-related non-cognitive factors

Summary: This study will examine the extent to which different aspects of spatial ability (i.e., navigation, object manipulation and visualisation), distilled from a battery of 16 tests, can predict STEM (science, technology, engineering and mathematics) educational outcomes, post-secondary degree choice and STEM-related non-cognitive factors, and the extent of their prediction after accounting for general cognitive ability. The project will leverage data from the longitudinal Twins Early Development Study (TEDS), including a comprehensive battery of spatial ability tests collected when the twins were 18 years old. We will employ factor analysis, linear and multiple regression within a structural equation model (SEM) framework to: a) produce latent variables for the three factors of spatial ability (as previously described in Malanchini et al., 2020a), b) analyse the association between these variables and key STEM-related education and life outcomes, c) examine the specific predictive ability of spatial skills beyond general cognitive ability. Finally, we will explore the aetiology driving the associations between spatial skills and educational and life outcomes by conducting bivariate and trivariate Cholesky decompositions. Background: STEM (science, technology, engineering and mathematics) education is becoming increasingly important in our increasingly technological world. For governments and businesses, higher student engagement in STEM represents an investment in human capital development; at the individual level, success in STEM is associated with a myriad of positive life outcomes, including income, well-being, and all-cause mortality (Goodman, 2019; Betts and Rose, 2001, Dreary, Hill and Gale, 2021). Given this, it is important to first understand the factors that predict both success in STEM education and one’s decision to pursue STEM careers—the basic knowledge necessary to increase overall engagement in STEM as well as reduce disparities in STEM outcomes. As with achievement in education more broadly, success in STEM subjects at school varies in the population. Research has shown that much of the individual variation in educational achievement can be accounted for by inherited DNA differences between individuals (see Malanchini, Rimfeld, Allegrini, Ritchie & Plomin, 2020 for a review). Both cognitive ability and non-cognitive factors (e.g., self-efficacy, motivation and mathematics anxiety) have been shown to contribute to the heritability of educational achievement, with shared genetic influences between the cognitive and non-cognitive factors explaining around 75% of achievement’s total heritability (Krapohl et al., 2014). Compared to the extant literature on educational achievement more broadly, however, there has been little genetically-informed work focusing specifically on STEM achievement and its predictors. A recent study exploring the genetic and environmental overlap between emotions, attitudes and performance (i.e. non-cognitive factors) in mathematics found a strong genetic overlap between these mathematics-related traits (Malanchini et al., 2020b). This provides a first line of evidence for delving deeper into the investigation of how students develop, select and modify STEM-related attitudes and experiences, partly based on their genetic predispositions. In addition to these non-cognitive factors, previous literature has shown that spatial ability (a domain of cognitive ability defined broadly as one’s ability to manipulate mental representations of objects) predicts STEM outcomes (Wai, Lubinsky and Penbow, 2009). However, theoretical discordances with regards to the architecture of the construct have stymied its adoption in psychological and educational longitudinal research. Specifically, it is unknown whether the different subdomains of spatial ability represent a single construct or several different ones. In an effort to elucidate this architecture, Malanchini et al. (2020c) employed factor analysis on a comprehensive novel battery of 16 spatial ability tests and found evidence for a hierarchical model in which 3 factors emerged (navigation, object manipulation and visualisation). All 3 factors were highly intercorrelated (mainly for genetic reasons) and loaded substantially onto a general factor. This unitary factor of spatial ability was found to be highly heritable at 84%. Critically, 45% of this genetic influence was independent of genetic influence on general cognitive ability. This finding supports previous work finding that spatial ability may be distinguishable, both phenotypically and genetically, from general cognitive ability (Rimfeld et al., 2017). In other words, spatial ability can be thought of as a cognitive skill in its own right, distinct from—and thus relatively untapped by—measures of general cognitive ability. The finding that spatial ability is characterized by a multi-factorial structure that appears to be distinct from general cognitive ability presents exciting new avenues of inquiry for understanding the role of spatial skills in propelling individuals on different STEM trajectories. Principally, it allows us to explore the heterogeneity (or lack thereof) in predictive value amongst spatial domains: does, for example, object manipulation predict certain STEM outcomes better than navigation? Furthermore, a paucity of research has examined the role of spatial ability for STEM outcomes within a developmental context. Fundamentally, no research to date has investigated the contribution of genetic and environmental factors to their association, not least with a comprehensive battery of spatial ability tests tapping multiple domains of spatial ability. The present study will leverage this unique, novel, and comprehensive battery of spatial tests alongside longitudinal data from the Twins Early Development Study (TEDS) to investigate the specific role of the three subdomains of spatial ability in predicting STEM outcomes (beyond general cognitive ability) within a developmental context, and to explore the aetiology underlying these associations. For comparison, we will also associate spatial ability with non-STEM educational outcomes (e.g. English and humanities grades).